Assessing the capacity of methylated DNA markers of cervical squamous cell carcinoma to discriminate oropharyngeal squamous cell carcinoma in human papillomavirus mediated disease

Abstract

Objective:

Early identification of human papillomavirus associated oropharyngeal squamous cell carcinoma (HPV(+)OPSCC) is challenging and novel biomarkers are needed. We hypothesized that a panel of methylated DNA markers (MDMs) found in HPV(+) cervical squamous cell carcinoma (CSCC) will have similar discrimination in HPV(+)OPSCC tissues.

Materials and Methods:

Formalin-fixed, paraffin-embedded tissues were obtained from patients with primary HPV(+)OPSCC or HPV(+)CSCC; control tissues included normal oropharynx palatine tonsil (NOP) and cervix (NCS). Using a methylation-specific polymerase chain reaction, 21 previously validated cervical MDMs were evaluated on tissue-extracted DNA. Discrimination between case and control cervical and oropharynx tissue was assessed using area under the curve (AUC).

Results:

34 HPV(+)OPSCC, 36 HPV(+)CSCC, 26 NOP, and 24 NCS patients met inclusion criteria. Within HPV(+)CSCC, 18/21 (86%) of MDMs achieved an AUC ≥ 0.9 and all MDMs exhibited better than chance classifications relative to control cervical tissue (all p<0.001). In contrast, within HPV(+)OPSCC only 5/21 (24%) MDMs achieved an AUC ≥ 0.90 but 19/21 (90%) exhibited better than chance classifications relative to control tonsil tissue (all p<0.001). Overall, 13/21 MDMs had statistically significant lower AUCs in the oropharyngeal cohort compared to the cervical cohort, and only 1 MDM exhibited a statistically significant increase in AUC.

Conclusions:

Previously validated MDMs exhibited robust performance in independent HPV(+)CSCC patients. However, most of these MDMs exhibited higher discrimination for HPV(+)CSCC than for HPV(+)OPSCC. This suggests that each SCC subtype requires a unique set of MDMs for optimal discrimination. Future studies are necessary to establish an MDM panel for HPV(+)OPSCC.

INTRODUCTION

Incidence of oropharyngeal squamous cell carcinoma (OPSCC) is increasing at a faster rate than other head and neck cancers in the United States¹, attributable to changes in human papillomavirus associated OPSCC (HPV(+)OPSCC) incidence². Indeed, HPV(+)OPSCC outpaced HPV(+) cervical squamous cell carcinoma (HPV(+)CSCC) in 2015² and is now the most common HPV associated cancer in the US³. While HPV vaccination could prevent over 90% of all HPV tumors, based on current vaccination rates, cases of HPV(+)OPSCC are not expected to plateau until 2045⁴. Identification of novel biomarkers to aid in early diagnosis or surveillance of HPV(+)OPSCC are lacking.

In general, as methods for treating cancer continue to become more sophisticated, it has become increasingly clear that earlier detection of tumors is critical. Particular emphasis has been placed on non-invasive or minimally invasive measures, which is of particular interest for HPV(+)OPSCC where no detectable premalignant lesion has been identified. Over the last 5-10 years, there has been a rapid increase in the investigation of tests which aim to identify substances shed from tumors into the blood stream. Most commonly, circulating tumor DNA (ctDNA) has been used to diagnose cancer via liquid biopsy⁵. Rettig et al. and Routman et al. separately investigated patients with HPV(+)OPSCC and each found an approximately 89% detection rate at diagnosis, which was associated with nodal status and lymphovascular invasion in patients who underwent surgery^{6, 7}. However, more than half of patients without nodal disease had undetectable ctDNA at presentation. Thus, while ctDNA may be useful for treatment selection and surveillance, its performance for detecting early HPV(+)OPSCC appears limited.

Studies of multiple cancers have examined the usefulness of other techniques, such as cell-free DNA (cfDNA) methylation^8-10, genetic mutations¹¹ and DNA fragmentation¹². Liu et al. targeted a panel of >100,000 key methylation regions in plasma cfDNA from 6,689 participants (2482 cancer, 4207 non-cancer)⁹. They found that targeted methylation analysis of cfDNA detected more than 50 cancer types across all stages, at a specificity of >99%. Tu et al. found that DNA methylation diagnostic biomarkers for CSCC could be successfully used to differentiate from normal cervical tissue¹³. This has similarly been found in pancreatic¹⁴, triple negative breast cancer¹⁵, prostate cancer¹⁶, and ovarian cancer¹⁷, among others. Despite the epidemic rise in the disease, this has not yet been done for HPV(+)OPSCC.

Oncogenic viral infections alter DNA methylation¹⁸ making this a particularly attractive candidate biomarker class for HPV(+)OPSCC¹⁹. In HPV mediated cancer, HPV protein E6 increases DNA methylation by inactivation of p53, a DNA methyltransferase 1 (DNMT1) inhibitor²⁰. Consequently, there has been increased interest in evaluating epigenetic differences between HPV-driven and non-HPV-driven cancers rather than differences between cancer and non-cancer²⁰. It is possible that HPV-mediated SCC may have similarities across subsites. As HPV mediates not only OPSCC, but also CSCC, we hypothesized that methylated DNA markers of HPV(+)CSCC would provide strong discrimination between HPV(+)OPSCC and NOP. We have previously validated a panel of methylated DNA markers (MDM) that discriminated between HPV(+)CSCC and benign cervical tissues with individual AUCs >0.90²¹. Here we sought to assess classification accuracy of these MDMs in HPV(+)OPSCC.

MATERIALS AND METHODS

Study population and patient selection

After institutional review board approval (IRB: 19-005075) we identified patients with HPV(+)CSCC, HPV(+)OPSCC, normal oropharyngeal palatine tonsil tissue (NOP), and normal uterine cervical tissue (NCS) treated at Mayo Clinic, Rochester, MN. HPV(+)OPSCC patients were identified through the Department of Otolaryngology Oropharynx REDCap Cancer database (1990-2019 at time of data retrieval). All tissues were formalin-fixed, paraffin-embedded (FFPE) and identified from an archive of residual clinical specimens in the Mayo Clinic Tissue Registry. Clinical history was obtained from the medical record and included age at diagnosis, sex at birth (male vs female), race (white vs non-white), tobacco and alcohol use, cervical Papanicolaou (Pap) test history (if relevant), and comorbidity status as defined using the Adult Comorbidity Evaluation-27 (ACE-27)²².

Patients were included if they had a primary (non-recurrent) tumor (OPSCC; CSCC) and normal histology (controls; NOP, NCS), no prior history of neoplastic or malignant disease of the target site (head and neck; uterine cervix), no exposure to chemotherapeutics within 1 year, no prior therapeutic radiation to the target region, no prior transplant, had clinical history available for review, sufficient target tissue available (>5mm), and were > 18 years of age. Normal cervical patients were excluded for any history of an abnormal Pap smear. Patients were considered HPV(+) if they had positive DNA in-situ hybridization for high-risk HPV (16, 18, 31, 33, or 51) or positive results on p16 staining (>70% diffuse nuclear and cytoplasmic staining). Patients were excluded if they declined Minnesota Research Authorization, if they had metastatic disease at presentation, if they presented with a synchronous primary tumor, or had a second primary active malignancy.

Tissue procurement and quantitative methylation specific PCR

2 mm FFPE tissue cores (marked by JJG) between 10 and 12 mm deep were macrodissected by the Biospecimens Accession and Processing Core lab from the area of interest within each block received from the Tissue Registry. The panel of methylated DNA markers (MDMs) was previously validated for HPV(+)CSCC²¹. These markers were then evaluated using DNA from independent FFPE HPV(+)OPSCC, HPV(+)CSCC, NOP, and NCS tissue using methylation-specific polymerase chain reaction (PCR). White blood cells (WBCs) were used as a background control.

Candidate MDMs were assayed by quantitative methylation specific PCR assays (qMSP), as previously published¹⁷. DNA was purified using the Qiagen (Valencia, CA) QIAmp FFPE DNA Tissue Kit (part#56404). Ten ng of sample DNA (per MDM) was bisulfite converted using the EZ-96 DNA Methylation kit (Zymo Research, Irvine CA) and amplified with primers recognizing the converted sequence using SYBR Green detection on Roche 480 LightCyclers (Roche, Basel Switzerland). Serially diluted universally methylated DNA samples were used as positive control standards, and negative controls included bisulfite converted and unconverted leukocyte-derived genomic DNA, and converted whole genome amplified (unmethylated) DNA. All laboratory staff were blinded to sample group information (cancer versus normal).

Statistical analysis

A target accrual of at least 31 patients per group was set to give 80% power for detecting an AUC of 0.85 given a null AUC of 0.7 while fixing alpha to 0.05. Up to 36 patients per group were included, where possible, to protect against potential assay failure or other sample dropout. Continuous variables were summarized as median (interquartile range), and discrete variables were summarized as frequency (percent). Each MDM was normalized to β-actin using generalized additive models (GAM) fit to the respective control group (NCS for HPV(+)CSCC; NOP for HPV(+)OPSCC). Discrimination between groups was assessed using AUC with corresponding 95% confidence intervals. A standard z-test was used to compute (unpaired) p-values for comparing performance between cervical and oropharyngeal tissue. Stratified AUCs were summarized by sex in oropharyngeal patients. The β-actin adjusted mean marker level across all 21 MDMs was used as a global measure of background signal in control patients and summarized for NCS and NOP separately. All analyses were performed in R (version 4.1; R Core Team). Two-sided P < 0.05 was considered statistically significant.

Data availability

The data generated in this study are available upon request from the corresponding author, pending approval from the Mayo Clinic IRB and Exact Sciences.

RESULTS

Patient demographics

36 HPV(+)CSCC, 24 NCS, 34 HPV(+)OPSCC, and 26 NOP patients met inclusion criteria. As shown in Table 1, HPV(+)OPSCC patients were older (median 57, IQR=48-62 years) than NOP patients (median 45, IQR=33-66 years) and had slightly increased alcohol and tobacco exposure. ACE-27 comorbidity scores were similar between the two groups. Males represented 88% of the HPV(+)OPSCC and 58% of NOP patients. When compared to HPV(+)CSCC, HPV(+)OPSCC patients were older, more often male, more likely to use alcohol and less likely to use tobacco. Tumors were identified earlier in HPV(+)CSCC, with 83% at stage I compared to 47% at stage I in HPV(+)OPSCC. More stage II tumors were found in HPV(+)OPSCC (44% vs 11% in HPV(+)CSCC), and stage III tumors were comparably represented (9% and 6% for HPV(+)OPSCC and HPV(+)CSCC, respectively).

Table 1:

Clinical characteristics of case and control patients providing tissues for methylation analysis.

	CSCC (N=36)	NCS (N=24)	OPSCC (N=34)	NOP (N=26)
Age
Median (Q1, Q3)	43 (37, 53)	44 (38, 55)	57 (48, 62)	45 (33, 66)
Female	36 (100.0%)	24 (100.0%)	4 (11.8%)	11 (42.3%)
Caucasian	26 (72.2%)	24 (100.0%)	33 (97.1%)	25 (96.2%)
Alcohol
N-Miss	16	1	1	2
No	11 (55.0%)	7 (30.4%)	5 (15.2%)	6 (25.0%)
Yes	9 (45.0%)	16 (69.6%)	28 (84.8%)	18 (75.0%)
Ever Tobacco Use
N-Miss	2	1	0	0
No	11 (32.4%)	11 (47.8%)	19 (55.9%)	17 (65.4%)
Yes	23 (67.6%)	12 (52.2%)	15 (44.1%)	9 (34.6%)
ACE-27
1	19 (52.8%)	5 (20.8%)	17 (50.0%)	8 (30.8%)
2	9 (25.0%)	15 (62.5%)	8 (23.5%)	8 (30.8%)
3	6 (16.7%)	3 (12.5%)	7 (20.6%)	8 (30.8%)
4	2 (5.6%)	1 (4.2%)	2 (5.9%)	2 (7.7%)
PAP Test Results
N-Miss	35	15	34	21
ASCUS	0 (0.0%)	0 (0.0%)	0	1 (20.0%)
Negative	1 (100.0%)	9 (100.0%)	0	4 (80.0%)
Cancer Stage
N-Miss	0	24	0	26
I	30 (83.3%)	0	16 (47.1%)	0
II	4 (11.1%)	0	15 (44.1%)	0
III	2 (5.6%)	0	3 (8.8%)	0
Cancer Tissue Purity
N-Miss	0	24	0	26
Median (Q1, Q3)	70 (60, 70)	NA	80 (80, 80)	NA

^aCSCC: HPV(+) Cervical Squamous Cell Carcinoma; NCS: Normal Cervical Tissue; OPSCC: HPV(+) Oropharyngeal Squamous Cell Carcinoma; NOP: Normal Oropharyngeal Palatine Tonsil; ACE-27: Adult Comorbidity Evaluation-27; PAP: Papanicolaou; ASCUS: Atypical Squamous Cells of Undetermined Significance; NA: non-applicable

MDM panel shows better discrimination between benign and malignant tissue in the cervical setting

The discrimination of 21 MDMs was evaluated in HPV(+)CSCC and HPV(+)OPSCC compared to their respective controls. AUCs and corresponding 95% confidence intervals are displayed in Table 2, and MDM intensity is visualized in Figure 1. Within HPV(+)CSCC, 18/21 (86%) MDMs achieved an AUC ≥ 0.90 and all MDMs exhibited better than chance classifications relative to control cervical tissue (all p<0.001). Within HPV(+)OPSCC, 5/21 (24%) achieved an AUC ≥ 0.90, 15/21 (71%) achieved an AUC ≥ 0.8, and 19/21 (90%) exhibited better than chance classifications relative to control tonsil tissue (all p<0.001). AUCs of 7 MDMs (LOC645323, TSPYL5, C1orf114, ZNF773, ZNF781.F, ARHGAP12 and ZNF69) were not significantly different between HPV(+)OPSCC and HPV(+)CSCC. The AUC for one MDM (GRIN2D) was significantly higher in HPV(+)OPSCC. The remaining 13 MDMs had lower AUCs in HPV(+)OPSCC compared to HPV(+)CSCC. Distributions of individual β-actin -adjusted MDMs by group are presented in Figure 2.

Table 2:

Discrimination between HPV(+) cancers and their respective control groups

	CSCC vs. NCS AUC (95% CI)	OPSCC vs. NOP AUC (95% CI)	Cervical vs. Oropharyngeal P-valuea
ZNF781.R	1 (1-1)	0.92 (0.86-0.99)	0.020
ZNF610	0.99 (0.98-1)	0.89 (0.8-0.97)	0.016
TBX15	0.99 (0.96-1)	0.82 (0.72-0.93)	0.003
MAX.chr6	0.99 (0.97-1)	0.86 (0.76-0.96)	0.010
TTYH1	0.99 (0.98-1)	0.82 (0.71-0.93)	0.002
ASCL1	0.98 (0.93-1)	0.86 (0.76-0.95)	0.019
NID2	0.98 (0.95-1)	0.86 (0.77-0.95)	0.021
ABCB1	0.97 (0.93-1)	0.74 (0.61-0.86)	0.001
MAX.chr9	0.97 (0.92-1)	0.82 (0.72-0.93)	0.011
LOC645323	0.95 (0.89-1)	0.88 (0.8-0.97)	0.195
NEUROG3	0.95 (0.89-1)	0.7 (0.57-0.83)	0.001
TMEM200C	0.95 (0.89-1)	0.54 (0.39-0.69)	< 0.001
TSPYL5	0.94 (0.87-1)	0.82 (0.72-0.93)	0.058
C1orf114	0.94 (0.86-1)	0.88 (0.8-0.97)	0.337
EMX1	0.93 (0.86-0.99)	0.77 (0.65-0.89)	0.019
ZNF773	0.93 (0.86-1)	0.9 (0.82-0.98)	0.570
ZNF781.F	0.92 (0.86-0.99)	0.79 (0.67-0.91)	0.058
ARHGAP12	0.91 (0.83-0.99)	0.9 (0.81-1)	0.922
ZNF69	0.89 (0.8-0.97)	0.95 (0.9-1)	0.202
VWC2	0.88 (0.79-0.97)	0.57 (0.42-0.72)	0.001
GRIN2D	0.72 (0.59-0.85)	0.91 (0.83-0.99)	0.019

^aTwo-sided

^bCSCC: HPV(+) Cervical Squamous Cell Carcinoma; NCS: Normal Cervical Tissue; OPSCC: HPV(+) Oropharyngeal Squamous Cell Carcinoma; NOP: Normal Oropharyngeal Palatine Tonsil; AUC: Area under the curve; CI: confidence interval

Figure 1. MDM values were elevated in HPV(+) tumor tissue compared to corresponding control tissue.

MDM intensity and clinical covariates for normal cervical tissue (NCS) and HPV(+)CSCC (Panel A) and normal oropharyngeal palatine tonsil (NOP) and HPV(+)OPSCC (Panel B). Each row is a clinical covariate/MDM, and each column is a unique patient sample. MDM intensities were calculated separately for Panel A and Panel B. Marker levels below the 95^th percentile in each control group were set to dark gray; increasing color intensity from yellow to red reflects increasing IQR standardized distance above this threshold. For continuous and ordinal clinical covariates, light pink reflects the lowest value and blue the highest. For binary clinical covariates, blue represents “yes,” and light pink represents “no.” Missing data is reflected with white boxes.

Figure 2. Most methylated DNA markers (MDMs) show better separation between benign and malignant tissues in cervical compared to oropharyngeal tissues.

DNA was extracted from FFPE tumor tissues (OPSCC, n=34; CSCC, n=36), normal oropharyngeal palatine tonsil (NOP, n=26) and normal cervical tissue (NCS, n=24). Methylation of selected DNA markers was determined using quantitative methylation specific PCR assays, β-actin adjusted marker levels among the groups are depicted by violin plots with embedded box plots representing minimum, Q1, median, Q3 and maximum values).

Sex differences in the cohorts had no obvious effect on marker performance

As these MDMs were discovered and previously validated in cervical tissue (all female) and oropharyngeal tissues were obtained primarily from males (Table 1), we performed stratified analyses to assess sex differences in MDM performance (Table 3). For most markers, we observed similar performance across sex, although statistical precision among females was very poor. Marker levels differed between NCS and NOP (Figure 3), with NOP patients exhibiting higher median methylation levels and greater variance compared to NCS, which may be explained by the origin of the panel or may indicate less underlying variability in cervical tissue methylation compared to tonsil.

Table 3:

Sex-stratified and overall AUCs with corresponding 95% confidence intervals.

	OPSCC Males vs. NOP Males AUC (n=45)	OPSCC Females vs. NOP Females AUC (n=15)	OPSCC vs. NOP Overall AUC
ZNF781.R	0.91 (0.83-0.99)	0.95 (0.85-1)	0.92 (0.86-0.99)
ZNF610	0.88 (0.78-0.98)	0.91 (0.72-1)	0.89 (0.8-0.97)
TBX15	0.86 (0.75-0.96)	0.7 (0.23-1)	0.82 (0.72-0.93)
MAX.chr6	0.85 (0.74-0.96)	0.91 (0.75-1)	0.86 (0.76-0.96)
TTYH1	0.83 (0.71-0.95)	0.61 (0.17-1)	0.82 (0.71-0.93)
ASCL1	0.86 (0.76-0.97)	0.73 (0.41-1)	0.86 (0.76-0.95)
NID2	0.89 (0.79-0.98)	0.73 (0.44-1)	0.86 (0.77-0.95)
ABCB1	0.74 (0.6-0.88)	0.64 (0.27-1)	0.74 (0.61-0.86)
MAX.chr9	0.87 (0.77-0.97)	0.61 (0.25-0.98)	0.82 (0.72-0.93)
LOC645323	0.87 (0.77-0.97)	0.86 (0.65-1)	0.88 (0.8-0.97)
NEUROG3	0.7 (0.55-0.85)	0.57 (0.17-0.96)	0.7 (0.57-0.83)
TMEM200C	0.52 (0.35-0.7)	0.73 (0.45-1)	0.54 (0.39-0.69)
TSPYL5	0.83 (0.71-0.95)	0.75 (0.5-1)	0.82 (0.72-0.93)
C1orf114	0.88 (0.79-0.98)	0.82 (0.54-1)	0.88 (0.8-0.97)
EMX1	0.8 (0.68-0.93)	0.59 (0.23-0.95)	0.77 (0.65-0.89)
ZNF773	0.87 (0.77-0.98)	1 (1-1)	0.9 (0.82-0.98)
ZNF781.F	0.8 (0.67-0.93)	0.77 (0.38-1)	0.79 (0.67-0.91)
ARHGAP12	0.9 (0.79-1)	0.93 (0.78-1)	0.9 (0.81-1)
ZNF69	0.98 (0.94-1)	0.77 (0.42-1)	0.95 (0.9-1)
VWC2	0.59 (0.42-0.75)	0.39 (0.06-0.71)	0.57 (0.42-0.72)
GRIN2D	0.92 (0.83-1)	0.89 (0.65-1)	0.91 (0.83-0.99)

^aCSCC: HPV(+) Cervical Squamous Cell Carcinoma; NCS: Normal Cervical Tissue; OPSCC: HPV(+) Oropharyngeal Squamous Cell Carcinoma; NOP: Normal Oropharyngeal Palatine Tonsil; AUC: Area under the curve; CI: confidence interval

Figure 3. Mean methylation levels were elevated and had greater variance in NOP compared to NCS tissues.

Violin plots (with embedded box plots displaying minimum, Q1, median, Q3 and maximum) show adjusted mean marker level across 21 MDMs in control patients (normal cervical tissue (NCS) and normal oropharyngeal palatine tonsil (NOP)).

DISCUSSION

In this study, we observed that methylation markers discovered and validated in cervical cancer achieved some success in discriminating between HPV(+)OPSCC and NOP. 90% of markers examined provided better than chance classification between these groups. Additionally, 5/21 (24%) markers exhibited excellent discriminatory potential (AUC estimates > 0.9). Although there were fewer females in the OPSCC group, stratified analysis revealed similar discrimination in males and females (Table 3). While some markers performed well, discriminating each cancer type from its respective control tissue, overall, the markers did not achieve the same degree of discrimination in HPV(+)OPSCC as in HPV(+)CSCC.

It has been suggested that HPV infection results in an early change in the carcinogenesis timeline for epithelial cells, leading to key alterations in the tumor microenvironment (TME)²³. Sartor et al demonstrated that HPV(+) cell lines had higher DNA methylation in genic and LINE-1 regions than HPV(−) SCC cell lines¹⁹. This is consistent with our observation that several MDMs (ZNF781.R, ZNF773, ARHGAP12, ZNF69, GRIN2D) from our HPV(+)CSCC panel achieved excellent discrimination in HPV(+)OPSCC. However, most markers still performed better in HPV(+)CSCC, motivating ongoing targeted discovery and validation testing for HPV(+)OPSCC. As stated above, the HPV infection event is the key factor in tumorigenesis for these subsites, however, differences in the epithelial structure (reticulated vs non), tumor microenvironment (immune privileged vs non), microbiome, carcinogen exposure, and other factors likely impact the choice of methylation targets between these subsites.

Demographics and other risk factors have been shown to be associated with altered levels of methylation. Using methylation sensitive digital PCR assessments, Lei et al. found that heavy alcohol consumption was strongly correlated with accelerated epigenetic aging²⁴. In addition, consistent with trends in SEER age-adjusted incidence rates for oropharynx and tonsil cancer by sex, which in 2019 were 5 per 100,000 (95% CI 4.9-5.2) for males and 0.9 per 100,000 for females (95% CI 0.9-1.0)²⁵, we had significantly more male patients in our HPV(+)OPSCC cohort than female. It is known that sex hormones, aging and smoking can influence the epigenetics seen in different cancer populations^26-28. Our study lacked power to assess the impact of these other modulators on the methylation patterns we observed in our cohorts, underscoring the need for further investigation into methylation differences between HPV(+)OPSCC and NOP tissues.

The majority of the MDMs evaluated in this study reside in genes that have been previously associated with other solid cancers by our group (Supplemental Table 1). The most common cancer association was esophageal squamous cell carcinoma (ESCC) where over half of the 21 current MDMs resided in genes that were identified as very discriminative hypermethylation events²⁹. Additionally, 13 of the 21 MDMs we identified were found in differentially methylated genes of multiple cancers including pancreatic, colorectal, gastric, lung, ovarian, and lymphoma. These cancers have widely different histologies, microenvironments and little or no HPV involvement. AUCs for OPSCC with the multiple-cancer MDMs ranged from 0.57 to 0.92. Conversely, 8 of the 21 MDMs were unique to CSCC and the performance of these in OPSCC ranged from AUC of 0.54 to 0.95. Generally, there does not appear to be a strong relationship between the site specificity or universality of these MDMs and their ability to discriminate between HPV(+)OPSCC and NOP.

This study has several important limitations. First, while cervical cancer and oropharyngeal cancer comprise the majority of HPV mediated squamous cell carcinoma, our study did not investigate other HPV mediated subsites. In addition, we used tonsillectomy specimens from adult patients with non-malignant disease to serve as our normal oropharyngeal controls. However, these tissue specimens may not represent normal tonsil as these tonsils were likely removed for infectious or inflammatory reasons and they may not be representative of base of tongue tonsillar tissue, as this subsite is less likely removed for benign disease. Our ability to distinguish between tissues of origin was limited as we used a validated panel of MDMs for cervical cancer across all sites (normal cervix, normal tonsil, HPV(+)OPSCC, HPV(+)CSCC) without a validated panel in HPV(+)OPSCC for comparison. Finally, we were not able to accrue 31 patients to each group, thus slightly decreasing our power to detect differences between case and control tissues.

CONCLUSION

Using a panel of MDMs previously validated for HPV(+)CSCC, we found excellent discrimination between independent HPV(+)CSCC and NCS and only moderate discrimination between HPV(+)OPSCC and NOP. One criterion for the selection of the markers in HPV(+)CSCC tissue was low background levels in normal cervical tissues. The lower discrimination observed in HPV(+)OPSCC is likely reflective of the increased background levels in methylation in normal tissue, rather than decreased methylation of these markers in cancerous tissue (Figures 2 and 3). Differences in methylation levels may be dependent on demographics, such as alcohol or tobacco exposure, age, or sex, among others. Despite known differences in the rates of HPV(+)OPSCC in males and females, we did not identify a difference in methylation markers based on sex for HPV(+)OPSCC in our analysis, although statistical precision among females was poor. Future studies investigating a validated MDM panel for HPV(+)OPSCC and HPV driven disease across sites is needed.

Highlights.

• Methylated DNA markers distinguished cervical cancer from normal cervical tissue.

• Cervical panel performed well but not optimally in HPV-associated oropharynx cancer.

• A panel targeted to oropharynx cancer may be predictive of cancer and site-specific

ABBREVIATIONS:

ACE-27

Adult Comorbidity Evaluation-27

AUC

Area under the curve

CSCC

Cervical squamous cell carcinoma

cfDNA

Cell-free DNA

ctDNA

Circulating tumor DNA

FFPE

Formalin-fixed, paraffin-embedded

HPV

Human papillomavirus

HPV(+)OPSCC

HPV-associated oropharyngeal squamous cell carcinoma

MDM

Methylated DNA marker

NCS

Normal cervical tissue

NOP

Normal oropharynx palatine tonsil

OPSCC

Oropharyngeal squamous cell carcinoma